What is voice search?
Voice search is a search started by talking. The assistant on a phone, the speaker in a kitchen, the system in a car and the earbud on a run all listen for a question, turn speech into text, run a search or ask a model, and read back an answer. The technology arrived in the early 2010s with phone assistants and spread with smart speakers; it is now being rebuilt around large language models, which makes the answers more conversational and the sources less visible.
Spoken queries differ from typed ones. They are longer, phrased as questions and conversational: where is the nearest studio that does product photography, rather than product photography near me. Many are local and immediate, about places, hours, directions and calls. And the response is usually a single answer, drawn from a featured snippet, a business profile, a knowledge panel or, increasingly, a model's own synthesis, with no list of alternatives to scroll.
It is a share of search, not a replacement for it. The predictions a decade ago that half of all searches would be spoken did not come true. What did happen is that a meaningful slice of local and question-based searches moved to voice, and the brands that answer those questions clearly are the ones that get read out loud.
Why it matters
Why voice search matters
Voice is winner-takes-all. A typed search shows ten results; a spoken one reads one. Being the source of that one answer is worth more than ranking third, and being absent is invisible. For local businesses the stakes are direct: a spoken request for a place nearby is answered from business profiles and maps, and the one with complete, accurate details gets the call.
It also rewards plain language. A page that answers a question in a clear sentence near the top is the page an assistant can read aloud. Jargon, buried answers and pages that make the reader hunt do not survive the conversion to speech. And it tests brand names: a name that cannot be said or is heard as something else does not get found by voice at all.
The arrival of conversational assistants has widened the point. When a person asks an assistant which agency to call, the answer is assembled from everything the model has read about the candidates. Voice is one more reason a brand needs to be described consistently and clearly everywhere.
In practice
How voice search works in practice
- 01
Optimizing for voice is mostly search optimization done with spoken questions in mind. The work starts with the questions customers actually ask, found in search data, in sales conversations and in the questions people type into a site's own search. Pages are built to answer them: a question as a heading, a direct answer of a sentence or two beneath it, then the detail. FAQ sections and the structured data that marks them up help assistants find the answer.
- 02
Local listings are completed and kept accurate, because that is where voice sends its local traffic. Pages load fast, because an assistant will not wait. The brand's name, products and locations are written the way they are said. And the result is tested the obvious way, by asking the devices and listening to what comes back. Measurement is indirect, since voice traffic is not reported separately, so teams watch featured snippet wins, business profile actions and the questions their pages rank for.
Watch out
What people get wrong about voice search
Building a strategy around a decade-old prediction that never arrived. Treating voice as a separate discipline with its own tricks, when it is served by the same clear content that serves everyone. Ignoring local listings, which is where most voice traffic actually goes. Writing pages as lists of questions stuffed with keywords, which reads badly aloud and on the page. And declaring it unmeasurable and therefore ignorable, when the proxies are available to anyone who looks.
The quieter error is forgetting the brand itself. A name that is awkward to say, a product line that sounds like a competitor's or a studio whose address is listed three different ways is losing voice searches before any content is written.
